Embed Size (px)
Citation preview
2
3
Problem Frontier
Difficulty
Building Architecture and Construction
Limits of Materials
Limits of Cognition and Coordination
Software Architecture and Construction
Early Days of Computing
4
Problem Frontier Now
Difficulty
Building Architecture and Construction
Limits of Materials
Limits of Cognition and Coordination
Software Architecture and Construction
Computing Today
5
Problem Frontier Now
Difficulty
Building Architecture and Construction
Limits of Materials
Limits of Cognition and Coordination
Software Architecture and Construction
Q: What’s a good design? A: One people can understand.
Q: What’s a good architecture? A: One that a bunch of people can understand in the same way.
6
Increasing Attention to Human Factors
Amrit, C., Daneva, M., & Damian, D. (2014). Information and Software Technology 56 (2014) 1537–1542
7
LDA SVD SVM Etc.
There’s not much chemistry going on here!
Jim Likes to mix things up, put them on alcohol flame See if they catch fire or (YES!) explode Knows nothing, cares nothing about chemistry
8 Marcus, G. & Davis, E. Eight (No, Nine!) Problems with Big Data. New York Times, April 6, 2014.
Photo: I, MikeGogulski
This may be very useful. This is not science.
10
Medical Research Analogy • Bleeding common medical practice • Late 18th century • Francois Joseph Victor Broussais
• Promoted bleeding of “affected organ” • Pierre-Charles-Alexandre Louis
• Actual data collection about outcomes • Bleeding is not such a great idea
• Clinical trial is important, is not enough! • Nature of disease processes
• Empiricism in software engineering: “clinical trial” • How do we progress better than trial and error?
11
Takeaways, Part 1 • Theories must take account of the primary constraints
• Cognition, communication, organizations
• Data mining, machine learning, prediction models don’t further our understanding of the phenomenon
• Clinical trials – seeing if a given technique works – does not build understanding
• What we really need: • Theories spanning human behavior (from individual to
community) and software • Empirical tests of hypotheses derived from theories
12
Context: Three Promising Theories
• Transparency • Signaling • Collective intelligence
13
Transparency and Signaling
• Transparency • “accurate observability, of an organization’s low-level
activities, routines, behaviors, output, and performance”*
+ Politics, finance, government, workplace + Performing in front of an audience + Accountability, coordination - May hurt creativity and streamlining - Information overload
*Bernstein, E. S. (2012). The transparency paradox a role for privacy in organizational learning and operational control. Administrative Science Quarterly, 57(2), 181-216.
14
GitHub: Transparency
• Open source hosting environment • 8 million users, 16 million repositories • Social media functionality • Transparency
Example: Information Overload
16
17
18
19
Transparency and Signaling
• Signals • Original idea from evolutionary biology • Visible clues that imply hidden quality • Types of signals
• Assessment: visible clue cannot be produced without hidden quality
• Conventional: meaning is agreed upon, will continue to exist only if enforced by norms
20
Stotting as Honest Signal
21
Avoid Colorful Snakes
22
Avoid Colorful Snakes
23
Conventional Signals
24
Conventional Signals – Trustworthy?
CareerExcuse.com
25
26
Norms, Enforcement
27
GitHub Signals
• Hiring, evaluating people • Deciding what to pay attention to • Deciding what to use • Deciding where to contribute • Deciding what contributions to accept
28
Early Results
29
Signaling Takeaways
• A way to make sense of information-rich environment
• Theory creates expectations about how various signals can work • How is signal produced? • How easy/hard is it to evaluate? • What norms influence how it is used?
30
Collective Intelligence
• c exists! • Groups perform well on one task tend tend
to perform well on others • This effect is much stronger than average
or highest intelligence in group
31
Evidence for c
Woolley, A. W., Chabris, C.F., Pentland, A., Hashmi, N., & Malonge, T.W. (2010). Evidence for a Collective Intelligence Factor in the Performance of Human Groups. Science, 330, Oct. 2010, 686-688.
c is predicted by • Social sensitivity • Equality of participation • Proportion of women
32
Collective Intelligence Genome
Malone, T.W., Laubacher, R., & Dellarocas, C. (2010). The collective intelligence genome. MIT Sloan Management Review, 51(3), pp. 20-31.
33
Basic Genes
• What • Create • Decide
• Who • Hierarchy • Crowd
• Why • Money • Love • Glory
• How • Collection • Contest • Collaboration • Group decision • Individual decision
Create
Decide
35
Collective Intelligence Takeaways
• Understanding the interplay among • Structure/process • Incentives • Staffing • Goal
• Effective collaboration systems can be designed
Q&A Genome Create Crowd
Collaborate Love & Glory
